RDFMap and GeoSmushing

Hi Chris,

I touched on this in the rdfGeo IRC meeting, but I thought I should bring it
up more fully.

In your example, you have:

<geom2d:Point>
 <map:srs resource="http://nurl.org/0/geography/SRSCatalog/wgs84">
 <geom2d:x>-123.817</geom2d:x>
 <geom2d:y>46.183</geom2d:y>
</geom2d:Point>

However this doesn't work if you wish to have a point described in 2
co-ordinate systems, you get:

<geom2d:Point>
 <map:srs resource="http://nurl.org/0/geography/SRSCatalog/wgs84">
 <geom2d:x>-123.817</geom2d:x>
 <geom2d:y>46.183</geom2d:y>
 <map:srs resource="urn:JimmysProjection">
 <geom2d:x>1</geom2d:x>
 <geom2d:y>1</geom2d:y>
</geom2d:Point>

Which is obviously unusable.  This is a problem as it means RDF tools will
not able to conclude 2 places are the same unless they also understand how
to map from one projection to another - they'll never be able to smush
simply on what they learn from the RDF.

For example, if I have the following RDF document:

<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:airport="http://www.daml.org/2001/10/html/airport-ont#"
xmlns:pos="http://www.w3.org/2003/01/geo/wgs84_pos#"
xmlns:k="http://opencyc.sourceforge.net/daml/cyc.daml#"
xmlns:pos2="urn:geo/wgs72_pos#">
 <rdf:Description>
  <foaf:name>Albergho Hostel</foaf:name>
  <pos:lat>-34.05331</pos:lat>
  <pos:long>23.37181</pos:long>
 </rdf:Description>
 <rdf:Description>
  <k:inLocation rdf:resource="http://jibbering.com/travels/places.rdf#ZA"/>
  <pos2:lat>-34.05334</pos2:lat>
  <pos2:long>23.37181</pos2:long>
 </rdf:Description>
 <rdf:Description>
  <pos:lat>-34.05331</pos:lat>
  <pos:long>23.37181</pos:long>
  <pos2:lat>-34.05334</pos2:lat>
  <pos2:long>23.37181</pos2:long>
 </rdf:Description>
 </rdf:RDF>

Then an RDF tool that was taught that lat/long combined make an unambigous
property could safely conclude that Albergho Hostel was in South Africa,
even though it doesn't know how to convert from WGS72 to WGS84.  This I
think is a very important feature.

http://esw.w3.org/topic/InterpretationProperties

discusses this some more, and explains why the
http://www.w3.org/2003/01/geo/wgs84_pos# approach of having a property for
each projection is likely a lot more useful then adding a projection to each
point (if you're going to model it like that, then you'll need a projection
for each lat/lon pair, not each point IMO)

Cheers,

Jim.

Received on Thursday, 17 April 2003 11:42:44 UTC